July 2022 - Pharo-vm - lists.pharo.org

Re: About global caches

by StÃ©phane Ducasse

Ok :) > On 26 Jul 2022, at 11:50, Guille Polito <guillermo.polito(a)inria.fr> wrote: > > Haha no global lookup, these are related to adaptive optimisations and inlining decisions :) > >> El 26 jul 2022, a las 11:31, stephane ducasse <stephane.ducasse(a)inria.fr <mailto:stephane.ducasse@inria.fr>> escribiÃ³: >> >> About global lookup caches >> >> @Nahuel Palumbo I think we should take a good look at these FIVE papers >> - Surgical precision JIT compilers [PLDI'14] >> https://dl.acm.org/doi/10.1145/2666356.2594316 <https://dl.acm.org/doi/10.1145/2666356.2594316> >> >> - Efficient late binding of dynamic function compositions >> https://dl.acm.org/doi/10.1145/3357766.3359543 <https://dl.acm.org/doi/10.1145/3357766.3359543> >> >> - Making collection operations optimal with aggressive JIT compilation >> https://dl.acm.org/doi/10.1145/3136000.3136002 <https://dl.acm.org/doi/10.1145/3136000.3136002> >> >> - Dynamic versus Static Optimization Techniques for Object-Oriented Languages >> https://onlinelibrary.wiley.com/doi/10.1002/j.1096-9942.1995.tb00017.x <https://onlinelibrary.wiley.com/doi/10.1002/j.1096-9942.1995.tb00017.x> >> >> - A Survey of Adaptive Optimization in Virtual Machines >> https://ieeexplore.ieee.org/document/1386662 <https://ieeexplore.ieee.org/document/1386662>_______________________________________________ >> Pharo-vm mailing list -- pharo-vm(a)lists.pharo.org <mailto:pharo-vm@lists.pharo.org> >> To unsubscribe send an email to pharo-vm-leave(a)lists.pharo.org <mailto:pharo-vm-leave@lists.pharo.org> > > _______________________________________________ > Pharo-vm mailing list -- pharo-vm(a)lists.pharo.org > To unsubscribe send an email to pharo-vm-leave(a)lists.pharo.org

July 26, 2022

Re: About global caches

by Guille Polito

Haha no global lookup, these are related to adaptive optimisations and inlining decisions :) > El 26 jul 2022, a las 11:31, stephane ducasse <stephane.ducasse(a)inria.fr> escribiÃ³: > > About global lookup caches > > @Nahuel Palumbo I think we should take a good look at these FIVE papers > - Surgical precision JIT compilers [PLDI'14] > https://dl.acm.org/doi/10.1145/2666356.2594316 <https://dl.acm.org/doi/10.1145/2666356.2594316> > > - Efficient late binding of dynamic function compositions > https://dl.acm.org/doi/10.1145/3357766.3359543 <https://dl.acm.org/doi/10.1145/3357766.3359543> > > - Making collection operations optimal with aggressive JIT compilation > https://dl.acm.org/doi/10.1145/3136000.3136002 <https://dl.acm.org/doi/10.1145/3136000.3136002> > > - Dynamic versus Static Optimization Techniques for Object-Oriented Languages > https://onlinelibrary.wiley.com/doi/10.1002/j.1096-9942.1995.tb00017.x <https://onlinelibrary.wiley.com/doi/10.1002/j.1096-9942.1995.tb00017.x> > > - A Survey of Adaptive Optimization in Virtual Machines > https://ieeexplore.ieee.org/document/1386662 <https://ieeexplore.ieee.org/document/1386662>_______________________________________________ > Pharo-vm mailing list -- pharo-vm(a)lists.pharo.org > To unsubscribe send an email to pharo-vm-leave(a)lists.pharo.org

July 26, 2022

About IR design: a design overview of javascript cores DFG IR

by stephane ducasse

A design overview of JavaScriptCores DFG IR <https://www.youtube.com/watch?v=EFRzBb9NYjM> https://www.youtube.com/watch?v=EFRzBb9NYjM S

July 26, 2022

About MoreVM'22

by stephane ducasse

> Hi! > > Guille and me went to MoreVMs <https://2022.programming-conference.org/home/MoreVMs-2022> event this week. > > We presented > > Interpreter Register Autolocalisation: Improving the performance of efficient interpreters > <autolocalisation.pdf> > > > > > And here there is a summary about the presentations we saw: > > - Less Is More: Merging AST Nodes To Optimize Interpreters > > This presentation shows an optimisation to interpreters merging AST nodes into supernodes, the same idea of superinstructions but at AST level. Itâs implemented on Truffle framework. Now the possible supernodes are selected manually, they have tools to help in this tasks. The results in presented benchmarks show better performance (reducing run time) than without supernodes. > > > - Towards Just-in-time and Language-agnostic Mutation Testing > > This work presents a Mutation Testing framework for languages built using Truffle framework. They offer common language-agnostic mutators and possibility to implement language specific mutators. After that, they perform Mutation Testing using Truffle API to determine the quality of a suite of unit tests. > Iâm not sure why they call âJIT" Mutation Testing. The abstract show the capability to mutate compiled code, take dynamic information at run-time and determine the utility of mutants for skip some of them as techniques with impact in mutant testing efficiency. > > > - Who You Gonna Call? A Case Study about the Call-Site Behaviour in Ruby-on-Rails Applications > > Sophie studied ruby on rails applications to see how many call sites are mono/poly/mega morphic with the objective of observing if they present phased behavior. Phased behavior means that the application behaviour changes at some point during its execution, in which moment it does a change of phase. > > Thing is that call site caches are usually global to the entire lifetime of the application. > However, it may happen that some call-site is monomorphic for some time, then it's monomorphic with another method for some other time. But a normal PIC will make a PIC with two entries instead of resetting the cache and creating monomorphic calls. > > Then she studied how phased behaviour would impact positively some applications. > They then manually marked the start of new phases and observed if the degree of polymorphism got effectively reduced. > > > - Torchy: A Tracing JIT Compiler for PyTorch > > Pytorch is a tool to script machine learning applications in Python that uses the GPU. > The problem of pytorch is that it eagerly performs calls to the GPU for every call in Python, which makes lots of useless roundtrips. > He presented a tracing compiler that creates linear execution traces that are sent all at once to the GPU to avoid such overhead.

July 26, 2022

About global caches

by stephane ducasse

About global lookup caches @Nahuel Palumbo I think we should take a good look at these FIVE papers - Surgical precision JIT compilers [PLDI'14] https://dl.acm.org/doi/10.1145/2666356.2594316 - Efficient late binding of dynamic function compositions https://dl.acm.org/doi/10.1145/3357766.3359543 - Making collection operations optimal with aggressive JIT compilation https://dl.acm.org/doi/10.1145/3136000.3136002 - Dynamic versus Static Optimization Techniques for Object-Oriented Languages https://onlinelibrary.wiley.com/doi/10.1002/j.1096-9942.1995.tb00017.x - A Survey of Adaptive Optimization in Virtual Machines https://ieeexplore.ieee.org/document/1386662

July 26, 2022

About SIMD [from nico]

by stephane ducasse

â¢ A couple of interesting readings I found today: - SIMD Instructions Considered Harmful https://www.sigarch.org/simd-instructions-considered-harmful/ RISC-V designers explain the disadvantages of fixed-size vector instructions - RISC-V Vector Instructions vs ARM and x86 SIMD https://medium.com/swlh/risc-v-vector-instructions-vs-arm-and-x86-simd-8c9b… Mentions some disadvantages of size-independent vector instructions that are not mentioned in the prior article â¢ Both of these talk about design of CPU instructions, but I think they also apply to bytecode design

July 26, 2022

About become: improvements

by stephane ducasse

Hi guys This list is silent because we forgot systematically to post to itâ¦ :) We will try to catch up. Here is a nice email from our internal academix list. A more detailed summary We reviewed NicoRâs blogpost on how to access the VM simulator https://thepharo.dev/2022/06/30/setting-up-your-environment-for-working-on-… <https://thepharo.dev/2022/06/30/setting-up-your-environment-for-working-on-…> We reviewed the `#become:` implementation. We have some notes to publish https://notes.inria.fr/rI0U--tWT1GP1c4O3hve_Q?edit <https://notes.inria.fr/rI0U--tWT1GP1c4O3hve_Q?edit> We profiled it with instruments and saw some performance bottlenecks. We were only able to do 35K become per second, which is not a lot. We did some little improvements to avoid those bottlenecks and improve performance by ~500x. Now we can do 2M become per second. Then Daniel showed us the primitive `#adoptInstance:` and we did a similar analysis. Finally, Nahuel showed his advances in the permanent graph selection, and we did some garbage collection benchmarks. We saw that moving most packages code (classes, methods, literals,...) can cut full GC pauses by ~half. Stef

July 26, 2022

test

by Guille Polito

subject

July 4, 2022